Lewis and colleagues: retrieval-augmented generation

What changed

Coupled a seq2seq generator to a dense Wikipedia index and compared sequence-level and token-level latent retrieval for generation.

Scientific scope

The record is the exact twelve-author NeurIPS 2020 paper. Claims remain bounded to the reported index, models, tasks and metrics; retrieval does not guarantee correct evidence use, current knowledge, factuality or provenance.

Evidence

The linked sources identify this contribution and its scope. WRS has not independently reconstructed the scientific claims. The record is the exact twelve-author NeurIPS 2020 paper. Claims remain bounded to the reported index, models, tasks and metrics; retrieval does not guarantee correct evidence use, current knowledge, factuality or provenance.

Contributors

• The named authors listed in the linked research record · SCIENCE: Complete source byline retained in the research record; this collective display does not merge person identities or measure labor shares.

• NeurIPS page for retrieval-augmented generation
• arXiv record for retrieval-augmented generation

Provisional WRS assessment, edition 2026-10-16.1. Independent expert calibration is pending. This grade does not certify correctness or assign ownership.

Full scientific record · Browse contribution grades · Request a correction